SwallowCare
Extending Dysphagia Care to the Home via AI Personalization
Company
SwallowCare × GT MS-HCI
My Role
Lead Product Designer (Team of 4)
Tools
Figma · Qualtrics · UserTesting · Gemini
Timeline
Aug 2025 – Present
Description
Bridging the gap between clinical visits for dysphagia patients through structured, personalized daily adherence powered by AI.
Context
Achieved an 87% task success rate across diet logging and exercise flows during usability testing without tutorials. We pivoted from generic gamification to an AI-powered experience to handle vague inputs and distinct care paths for elderly users. Now moving to seed funding with sponsor Dr. Rinki Varindani Desai.
The System Failure in Dysphagia Care:
Complexity Meets Scarcity
Dysphagia is difficulty in swallowing. Our research spanning 11 interviews, 27 surveys, and 5 diary studies exposed a critical systemic mismatch. While the 50+ patient population surges, SLP supply remains stagnant. Clinicians are forced to rely on generic handouts that fail diverse needs. While all users need diet adjustment and symptom logging, stroke survivors require high-intensity rehabilitation and ALS patients need progressive compensation.
Competitive analysis confirmed patients are forced into a fragmented void of disconnected apps. Analysis of 1,129 patient social media interactions validated that this fragmentation creates a 'Digital Barrier,' driving anxiety and regression. We aim to replace the void with a 'One-Size-Fits-One' companion that adapts to the distinct clinical trajectory of every patient.
Plot Twist 1:
Why We Killed Gamification
Dysphagia care requires daily effort on diet and exercises. We initially hypothesized that 'Streaks' and an animated character would drive adherence through engagement. However, testing with patients and experts revealed a critical flaw: the abstract mascot caused confusion, while gamified mechanics triggered performance anxiety rather than motivation.
We pivoted to a clarity-first model, replacing the character with real-person video demos. Similarly, 'Swallow Score', based on validated EAT-10 assessment scale, replaced the abstract points and streaks, allowing patients to track their longitudinal progress quantitatively.
Plot Twist 2:
From Generic Exploration to Contextual Relevance
Our initial design relied on standard explorative libraries and comprehensive logging. However, think-aloud sessions revealed that 'nice-to-have' metrics like portion sizes and calories created friction without adding safety value. Broad libraries overwhelmed our elderly demographic. Critical answers like "Is this safe?" or "What exercise helps right now?" were buried under layers of irrelevant screens and information.
The Structural Pivot: We prioritized Relevance. We stripped the logging flow of calories to focus purely on Texture Safety. We also restricted the library to only display content matching the patient's specific longitudinal data.
The Interaction Pivot: To handle active inquiries without forcing users to search, we designed for Flexible AI Input. Instead of navigating deep menus to check a dish or find a symptom-relief exercise, users simply scan a physical menu or verbally describe their condition. The AI parses this 'messy' input and serves the single relevant clinical result."
The Solution: One-Size-Fits-One Ecosystem
Tailored Plan & Video Guidance: Operationalizing Adherence
Patients typically leave clinics with confusing paper handouts, leading to low adherence. Static diagrams fail to convey complex throat maneuvers safely. We replaced the stack of papers with a daily checklist and real-person video guides. This removes decision paralysis and ensures clinical fidelity, empowering patients to practice correctly without direct supervision.
The Safety Loop: Prediction & Verification
Eating carries two critical risks: choosing the wrong food and failing to track reactions. Standard apps ignore this cycle. Our AI Assistant answers "Is this safe?" By analyzing menus against IDDSI standards, the assistant proactively filters out high-risk textures, preventing accidents before they happen. The Diet Log answers "Was it safe?", capturing immediate reactions to build a rigorous clinical safety profile.
Quantifying Recovery: Clinical & Social Visibility
Recovery is often invisible, leaving providers without data between visits. We embedded the EAT-10 scale to track "Swallow Scores." Together with symptom and adherence logged, SwallowCare generates provider-ready reports weekly. Beyond the metrics, a lightweight community layer connects patients on similar trajectories, breaking the isolation of dysphagia.
Design System: Engineered for Cognitive Clarity
We established a strict Accessibility-First design language to support aging vision. Atkinson Hyperlegible, a typeface developed by the Braille Institute, and a 18px base size maximized readability. Beyond branding, colors served as a functional traffic-light system. 'Sage' for safety and 'Amber' for caution allow patients to assess food risks instantly, with all UI components meeting WCAG 2.1 contrast standards.
Impact & Reflection
Validated Success & Next Steps
Our prototype achieved an 87% task success rate on critical flows. Following final iterations based on testing insights, the project is now moving into development, to be followed by large-scale usability trials and a seed funding round with our clinical sponsor.
The Benchmarking Trap
Looking back, I initially designed for calorie counting because I was primed by standard fitness apps. I learned that benchmarking against the mass market can be a trap when designing for niche clinical needs. Sometimes the 'standard' pattern is not the 'right' pattern.
The AI Paradox
Our elderly users showed a surprisingly positive attitude toward AI. While we leverage its flexibility for handling complex inputs, the risk of hallucination remains a safety barrier. To mitigate this, our sponsor is currently training a domain-specific LLM on proprietary clinical data to ensure safety before scaling.
Credit
Team Member
Riley Liu, Raj Sureka, Emily Jeong
External Partner
Dr. Rinki Varindani Desai
Special Thanks
NFOSD, Ed Steger, Dr. Carrie Bruce, Kiki Marlam, and everyone participated in testing sessions.